Fig. 11From: Recognizing sports activities from video frames using deformable convolution and adaptive multiscale featuresThree sports-related action recognition datasets with varying numbers of characteristics are used to evaluate the proposed technique against existing alternatives. From what can be seen in the graph, the suggested model has a rather narrow set of requirements. However, when the number of features is decreased, EfficentNet and RegNetY display no variance in accuracyBack to article page